How an NPS Score Got Gamed by a Poorly Designed Incentive

A support leadership team tied part of quarterly bonuses to the team's average post-ticket NPS score. The logic seemed sound — reward good service, measure it with the metric everyone already tracks. Within a month, average NPS on closed tickets climbed from 41 to 58. Leadership was thrilled. Then churn for the quarter came in slightly worse than the prior one, which shouldn't happen if satisfaction genuinely improved.

How the gaming actually happened

Nobody falsified anything. Support agents, entirely rationally, started explicitly asking customers for a high score at the end of calls — "if you could rate us a 9 or 10 on the survey you'll get..." — and started being more selective about which tickets got closed before the survey triggered, closing the easy wins promptly and letting the genuinely hard, unresolved tickets sit open longer so they wouldn't trigger a survey in a frustrated state.

The moment a metric becomes someone's income, the metric stops measuring the thing it was built to measure.

What we changed

• Removed the direct bonus tie to NPS and replaced it with a broader scorecard including resolution time and reopened-ticket rate.
• Triggered the survey on a fixed delay after ticket closure rather than immediately, reducing the ability to time it.
• Cross-checked NPS trend against churn and reopened-ticket rate monthly as a sanity check against gaming.
• Banned scripted asks for a specific numeric score, which is against most survey platforms' terms anyway.

NPS wasn't a bad metric to track. It was a bad metric to attach a direct financial incentive to without a second metric watching for gaming. Any score a human can influence in the moment of asking for it will eventually get gamed if there's money riding on the number.